Search papers, labs, and topics across Lattice.
Stargazer is introduced as a scalable benchmark environment for evaluating AI agents on physics-grounded model-fitting tasks using radial-velocity time series data. The environment comprises 120 tasks of varying difficulty, including real archival cases, to assess agents' ability to recover correct physical system parameters. Evaluation of eight frontier agents revealed a disconnect between statistical fit and adherence to physical constraints, with agents often failing to recover correct parameters despite good statistical fits, even with increased compute.
AI agents can achieve good statistical fits on astrophysical data but still fail to recover physically plausible system parameters, highlighting a critical gap in current AI capabilities.
The rise of autonomous AI agents suggests that dynamic benchmark environments with built-in feedback on scientifically grounded tasks are needed to evaluate the capabilities of these agents in research work. We introduce Stargazer, a scalable environment for evaluating AI agents on dynamic, iterative physics-grounded model-fitting tasks using inference on radial-velocity (RV) time series data. Stargazer comprises 120 tasks across three difficulty tiers, including 20 real archival cases, covering diverse scenarios ranging from high-SNR single-planet systems to complex multi-planetary configurations requiring involved low-SNR analysis. Our evaluation of eight frontier agents reveals a gap between numerical optimization and adherence to physical constraints: although agents often achieve a good statistical fit, they frequently fail to recover correct physical system parameters, a limitation that persists even when agents are equipped with vanilla skills. Furthermore, increasing test-time compute yields only marginal gains, with excessive token usage often reflecting recursive failure loops rather than meaningful exploration. Stargazer presents an opportunity to train, evaluate, scaffold, and scale strategies on a model-fitting problem of practical research relevance today. Our methodology to design a simulation-driven environment for AI agents presumably generalizes to many other model-fitting problems across scientific domains. Source code and the project website are available at https://github.com/Gudmorning2025/Stargazer and https://gudmorning2025.github.io/Stargazer, respectively.
